Skip to content

User Interview Synthesiser

Example prompt: "When a new Fathom recording tagged 'user-research' shows up in our team library, pull the transcript, pick out the themes, pain points, and the best quotes, post a short summary to our #product-research Slack channel, and append the highlights to our 'User Research' page in Notion."

The Problem

User research interviews are the highest-signal input a product team gets, and they are also the easiest to lose. A discovery call gets recorded, somebody promises to "write it up", the recording sits in Fathom for two weeks, and by the time we open it the urgency has faded and the team's mental model has moved on without the input. The PM who runs the interviews knows what was said; the engineers and designers who need to hear it second-hand get nothing more than a Slack paraphrase. Over a quarter the team makes decisions on a fraction of the evidence the calls actually produced.

How GloriaMundo Solves It

We build a workflow triggered when a new recording tagged "user-research" appears in our Fathom library. An integration step pulls the call's transcript and metadata. An LLM step reads the transcript end-to-end and produces a structured note — the interviewee's context in one paragraph, the three or four most prominent themes with a one-line summary each, the specific pain points raised, and four to six quotable verbatim passages with timestamps. A conditional step checks the LLM's confidence — if the transcript is unusable (bad audio, accidentally truncated, off-topic chat) the workflow flags it for the PM and stops; otherwise it continues. An integration step appends the structured note as a new dated section in the "User Research" Notion page. A second integration step posts a short summary to the #product-research Slack channel with a link to the full Notion entry. Glass Box preview shows the extracted themes and the draft Slack summary before anything is posted.

Example Workflow Steps

  1. Trigger (webhook): Fires when a new Fathom recording with the tag "user-research" finishes processing.
  2. Step 1 (integration): Pull the transcript, participants, and call metadata from Fathom.
  3. Step 2 (LLM): Extract a one-paragraph interviewee context, themes, pain points, and quotable verbatims with timestamps.
  4. Step 3 (conditional): If the transcript is unusable, post a "this one needs a human" note to the PM and stop; otherwise continue.
  5. Step 4 (integration): Append a dated section to the "User Research" page in Notion containing the structured note.
  6. Step 5 (integration): Post a short summary to #product-research in Slack with the themes, the top quote, and a link to the Notion entry.

Integrations Used

  • Fathom — source of the recorded interview and its transcript
  • Notion — long-form home of the structured research notes, browseable across quarters
  • Slack — where the team sees the headlines and gets nudged to read the full note

Who This Is For

Product managers and researchers at scale-ups running 4-10 user interviews a month, where the team relies on calls being written up promptly but the write-ups slip when the week gets busy. Especially useful when designers, engineers and PMs are in different time zones and the recording-plus-summary is the team's only practical shared input.

Time & Cost Saved

A careful write-up of a 45-minute interview takes 30-60 minutes of focused PM time and frequently does not happen at all in the week the interview was recorded. This workflow produces the structured note within minutes of the recording finishing, and the PM's time goes into the decisions the note surfaces rather than the transcription work.